youtube-dl/youtube_dl/extractor/roosterteeth.py

138 lines
5.6 KiB
Python
Raw Permalink Normal View History

2016-06-22 07:58:42 +00:00
# coding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
from ..compat import (
compat_HTTPError,
compat_str,
)
2016-06-22 07:58:42 +00:00
from ..utils import (
ExtractorError,
2016-07-09 18:28:28 +00:00
int_or_none,
str_or_none,
2016-06-22 07:58:42 +00:00
urlencode_postdata,
)
class RoosterTeethIE(InfoExtractor):
_VALID_URL = r'https?://(?:.+?\.)?roosterteeth\.com/(?:episode|watch)/(?P<id>[^/?#&]+)'
2016-06-22 07:58:42 +00:00
_NETRC_MACHINE = 'roosterteeth'
_TESTS = [{
'url': 'http://roosterteeth.com/episode/million-dollars-but-season-2-million-dollars-but-the-game-announcement',
2016-07-09 18:28:28 +00:00
'md5': 'e2bd7764732d785ef797700a2489f212',
2016-06-22 07:58:42 +00:00
'info_dict': {
'id': '9156',
2016-07-09 18:28:28 +00:00
'display_id': 'million-dollars-but-season-2-million-dollars-but-the-game-announcement',
2016-06-22 07:58:42 +00:00
'ext': 'mp4',
'title': 'Million Dollars, But... The Game Announcement',
'description': 'md5:168a54b40e228e79f4ddb141e89fe4f5',
'thumbnail': r're:^https?://.*\.png$',
2016-06-22 07:58:42 +00:00
'series': 'Million Dollars, But...',
'episode': 'Million Dollars, But... The Game Announcement',
},
}, {
'url': 'http://achievementhunter.roosterteeth.com/episode/off-topic-the-achievement-hunter-podcast-2016-i-didn-t-think-it-would-pass-31',
'only_matching': True,
}, {
'url': 'http://funhaus.roosterteeth.com/episode/funhaus-shorts-2016-austin-sucks-funhaus-shorts',
'only_matching': True,
}, {
'url': 'http://screwattack.roosterteeth.com/episode/death-battle-season-3-mewtwo-vs-shadow',
'only_matching': True,
}, {
'url': 'http://theknow.roosterteeth.com/episode/the-know-game-news-season-1-boring-steam-sales-are-better',
'only_matching': True,
2016-07-09 18:28:28 +00:00
}, {
# only available for FIRST members
'url': 'http://roosterteeth.com/episode/rt-docs-the-world-s-greatest-head-massage-the-world-s-greatest-head-massage-an-asmr-journey-part-one',
'only_matching': True,
}, {
'url': 'https://roosterteeth.com/watch/million-dollars-but-season-2-million-dollars-but-the-game-announcement',
'only_matching': True,
2016-06-22 07:58:42 +00:00
}]
_EPISODE_BASE_URL = 'https://svod-be.roosterteeth.com/api/v1/episodes/'
2016-06-22 07:58:42 +00:00
def _login(self):
username, password = self._get_login_info()
2016-07-09 18:28:28 +00:00
if username is None:
return
2016-06-22 07:58:42 +00:00
try:
self._download_json(
'https://auth.roosterteeth.com/oauth/token',
None, 'Logging in', data=urlencode_postdata({
'client_id': '4338d2b4bdc8db1239360f28e72f0d9ddb1fd01e7a38fbb07b4b1f4ba4564cc5',
'grant_type': 'password',
'username': username,
'password': password,
}))
except ExtractorError as e:
msg = 'Unable to login'
if isinstance(e.cause, compat_HTTPError) and e.cause.code == 401:
resp = self._parse_json(e.cause.read().decode(), None, fatal=False)
if resp:
error = resp.get('extra_info') or resp.get('error_description') or resp.get('error')
if error:
msg += ': ' + error
self.report_warning(msg)
2016-06-22 07:58:42 +00:00
def _real_initialize(self):
if self._get_cookies(self._EPISODE_BASE_URL).get('rt_access_token'):
return
2016-06-22 07:58:42 +00:00
self._login()
def _real_extract(self, url):
2016-07-09 18:28:28 +00:00
display_id = self._match_id(url)
api_episode_url = self._EPISODE_BASE_URL + display_id
try:
m3u8_url = self._download_json(
api_episode_url + '/videos', display_id,
'Downloading video JSON metadata')['data'][0]['attributes']['url']
except ExtractorError as e:
if isinstance(e.cause, compat_HTTPError) and e.cause.code == 403:
if self._parse_json(e.cause.read().decode(), display_id).get('access') is False:
self.raise_login_required(
'%s is only available for FIRST members' % display_id)
raise
2016-07-09 18:28:28 +00:00
formats = self._extract_m3u8_formats(
m3u8_url, display_id, 'mp4', 'm3u8_native', m3u8_id='hls')
2016-06-22 07:58:42 +00:00
self._sort_formats(formats)
episode = self._download_json(
api_episode_url, display_id,
'Downloading episode JSON metadata')['data'][0]
attributes = episode['attributes']
title = attributes.get('title') or attributes['display_title']
video_id = compat_str(episode['id'])
thumbnails = []
for image in episode.get('included', {}).get('images', []):
if image.get('type') == 'episode_image':
img_attributes = image.get('attributes') or {}
for k in ('thumb', 'small', 'medium', 'large'):
img_url = img_attributes.get(k)
if img_url:
thumbnails.append({
'id': k,
'url': img_url,
})
2016-07-09 18:28:28 +00:00
2016-06-22 07:58:42 +00:00
return {
2016-07-09 18:28:28 +00:00
'id': video_id,
'display_id': display_id,
2016-06-22 07:58:42 +00:00
'title': title,
'description': attributes.get('description') or attributes.get('caption'),
'thumbnails': thumbnails,
'series': attributes.get('show_title'),
'season_number': int_or_none(attributes.get('season_number')),
'season_id': attributes.get('season_id'),
'episode': title,
'episode_number': int_or_none(attributes.get('number')),
'episode_id': str_or_none(episode.get('uuid')),
2016-07-09 18:28:28 +00:00
'formats': formats,
'channel_id': attributes.get('channel_id'),
'duration': int_or_none(attributes.get('length')),
2016-06-22 07:58:42 +00:00
}